← Back to PRs

#18896: CLI: add vLLM configure command

by franciscojavierarceo open 2026-02-17 05:15 View on GitHub →
docs cli scripts commands size: M
## Summary - add `openclaw models configure vllm` command (interactive + flags) - write vLLM provider config and default model selection - fix `pnpm openclaw --` arg passthrough in dev runner ## Testing - `pnpm test src/commands/models.configure-vllm.e2e.test.ts` (fails: test config excludes `*.e2e.test.ts`) - `pnpm test:e2e -- src/commands/models.configure-vllm.e2e.test.ts` (runs full e2e suite; unrelated failures) - `pnpm test src/cli/models-cli.test.ts` <!-- greptile_comment --> <h3>Greptile Summary</h3> Adds `openclaw models configure vllm` command to streamline vLLM provider setup by prompting for configuration (or accepting CLI flags) and updating both auth profiles and config in one step. - Refactored `promptAndConfigureVllm` into reusable helpers (`normalizeVllmBaseUrl`, `buildVllmModelDefinition`, `applyVllmProviderConfig`) to support both interactive and non-interactive workflows - New command handles TTY detection and falls back to requiring `--api-key` and `--model-id` flags in non-interactive environments - Updates default model and adds to allowlist after configuration - Test coverage includes e2e test validating config updates, auth profile creation, and model preservation <h3>Confidence Score: 5/5</h3> - Safe to merge - well-structured refactor with solid test coverage - Clean implementation following existing patterns, good separation of concerns with extracted helper functions, comprehensive test coverage for both interactive and non-interactive paths, and proper error handling for edge cases - No files require special attention <sub>Last reviewed commit: e659b26</sub> <!-- greptile_other_comments_section --> <sub>(2/5) Greptile learns from your feedback when you react with thumbs up/down!</sub> <!-- /greptile_comment -->

Most Similar PRs